convenience, accessibility, and the demand for fast food · convenience, accessibility, and the...

Journal ofAgricultural and Resource Economics 26(1):58-74Copyright 2001 Western Agricultural Economics Association

Convenience, Accessibility, andthe Demand for Fast Food

Mark D. Jekanowski, James K. Binkley,and James Eales

This study explores the growth in demand for fast food. A distinguishing characteristicof fast food is its convenience; in today's pervasive marketplace, consumers need nottravel far to find a fast food outlet. This greater availability translates into a decreasein the full price of obtaining a meal, which contributes to greater consumption.Market-level data are used to estimate demand equations in two time periods, incor-porating changes in availability as well as prices, income, and various demographiccharacteristics. Our findings show that greater availability has led to increasedconsumption. Failure to account for these types of marketplace changes could lead toincorrect inferences regarding the factors responsible for the industry growth.

Key words: convenience, fast food, supplier-induced demand, travel costs

Introduction

U.S. food consumption patterns have altered dramatically over the past several decades,with the greatest change being the rise in consumer expenditures on meals preparedoutside the home (food away from home, or FAFH hereafter). In 1997, FAFH accountedfor about 45% of total food expenditures, up from approximately 26% in 1960 (Manchesterand Clauson). Several consumer demand studies have examined FAFH expenditures(e.g., Kinsey; Redman; Lee and Brown; Sexauer; Prochaska and Schrimper; Yen; Byrne,Capps, and Saha 1996,1998). Nearly all earlier studies have emphasized the importanceof household time and the rising value of convenience as factors driving FAFH con-sumption. Hence, variables measuring time costs, such as hours worked and female laborforce participation, are typically included in estimated models, along with income anddemographic characteristics. Convenience measures are often found to be importantdeterminants of food consumption patterns.

Symptomatic of the importance of convenience is a dramatic change within the FAFHmarket: the share of away-from-home food expenditures spent on fast food has beenincreasing steadily for several decades. In 1967, fast food accounted for 14.3% of totalaway-from-home food expenditures, and by 1999 it reached 35.5% (U.S. Department ofAgriculture/Economic Research Service). Despite its growing importance, fast food hasreceived little detailed attention in the academic literature, being considered in but afew of the FAFH studies (e.g., McCracken and Brandt; Brown; Ekelund and Watson;Hiemstra and Kim; Byrne, Capps, and Saha 1998).

Mark D. Jekanowski is a senior consultant with Sparks Companies, Inc., McLean, Virginia. James K. Binkley is professor,and James Eales is associate professor, both in the Department of Agricultural Economics, Purdue University. The authorswish to thank the JARE editor and two anonymous reviewers for helpful comments.

Review coordinated by Gary D. Thompson.

Convenience, Accessibility, and the Demandfor Fast Food 59

In past research examining demand for FAFH, "convenience" has mainly been definedas arising from the time saved by avoiding meal preparation. Consequently, emphasishas been placed on factors such as spouse employment and imputed time values. Whilethis view is generally sufficient for FAFH as a whole, it is somewhat limited in thespecific case of fast food. What distinguishes fast food from other types of FAFH is thatit is indeed fast-near immediate service, providing a consistent and popular product.Because of the standardized menu and consistent quality, only minimal time need bespent obtaining product information. But these inherent advantages of fast food are oflittle consequence unless the food is easily available. Consumers will not travel far for"reasonably good" food, nor will they drive 15 minutes to save 10 minutes in servicetime.1 Thus a third characteristic of fast food is critical: outlets are generally in easilyreached, nearby locations.

Accessibility is the leitmotif of industry strategy, as is apparent in the emphasis onmarket penetration. The National Restaurant Association, in its trade publicationRestaurants USA, notes:

Operators recognize consumers' need for convenience. Unit expansion continues tobe a key component of rapid growth in the limited-service segment, and higher unitcounts translate into greater consumer convenience, which in turn drives sales(Restaurants USA, December 1996, p. 13).

McDonald's, the fast food industry leader, avows a proactive stance that might bedescribed as an "in-your-face" strategy. As articulated in the corporation's 1994 annualreport:

McDonald's wants to have a site wherever people live, work, play, or gather. OurConvenience Strategy is to monitor the changing lifestyles of consumers and interceptthem at every turn. As we expand customer convenience, we gain market share(McDonald's Corporation USA, 1994 Annual Report, p. 8).

The current industry emphasis on "satellite" restaurants reflects this strategy. Thesesmaller, limited-menu, lower-volume units have considerably lower construction andoperating costs, and tend to be placed on leased property, further reducing the cost ofexpansion. They are often joined with convenience stores and gasoline stations, and insome cases are placed inside large retail stores.

In this study, we focus exclusively on fast food, and thus we emphasize the role ofmarket penetration as an explanatory factor, while considering prices, income, and othervariables traditionally included in demand studies. In order to develop an accessibilitymeasure and measures of price, we examine consumption across markets. With fewexceptions, previous research in FAFH demand has been based on household surveydata.

In the next section, we illustrate how availability can be modeled as a component ofthe full price e faced by consumers. The empirical analysis that follows reveals this aspectof convenience to be a key factor in explaining the long-term increase in meals preparedoutside the home, especially fast food.

1 Hiemstra and Kim found the average travel time to fast food outlets was 11 minutes, compared to 14 minutes for rest-aurants.

Jekanowski, Binkley, and Eales

Journal ofAgricultural and Resource Economics

Relationship Between Accessibilityand Quantity Demanded

Much of household production theory, on which convenience arguments are founded, issimply a monetization ofnonmarket costs incurred by the consumer. A basic premise is

that along with goods and services, time is incorporated into the utility-maximizationproblem (Becker). Consumers are concerned not only with the retail price of a product,

but also with the time costs incurred when purchasing and consuming the product. The

"full price" (P) is the sum of these two components, i.e., P = p + vt, where p is the retail

price paid at the counter, t is the time necessary to complete the transaction, and v is the

consumer value of time. Fast food suppliers emphasize minimization of time costs, that

is, the maximization of convenience.An important component of time costs is the time spent traveling to the retail outlet,

which is a direct function of distance. Construction of new outlets can decrease the

distance a consumer must travel, thereby lowering the full price of the product andincreasing the frequency of purchase. Crafton employed a model based on time costs to

describe price differences between convenience stores and supermarkets, but he focusedon time spent inside the establishment, holding travel distance constant. We focus instead

on travel time, which for fast food is perhaps the largest component of time costs. The

following illustration shows the effect of travel time on quantity demanded.In figure 1 we consider two consumers with identical prefereences and equal income

levels, but located at different distances from a retail outlet. To purchase a fast food

meal, consumer 1 must travel distance r1 to the outlet at a cost per unit of distance c,

making the full price of the product P =p + cr1. Consumer 2 resides further from the out-

let, so must travel distance r2, where r2 > r1. Assuming the same retail price p and cost

per unit of distance c, consumer 2 will always face a higher full price than consumer 1,

i.e., P = p + cr2 > P = p + cr1. Because the full price determines quantity demanded, con-

sumer 1 will demand Q1, and consumer 2 will demand Q2, where Q2 < Q1. If travel costswere absent, each consumer would demand quantity Q0. In an empirical setting, ignoring

travel costs could lead to the appearance of unequal preferences, as consumer 1 will

always demand more than consumer 2 at any given retail price. Similarly, changes in

travel costs over time could be improperly attributed to a shift in demand, rather than

a movement along the demand curve.If a new outlet is constructed closer to either consumer, r for that consumer will fall,

resulting in greater quantity demanded. Thus, r, the average distance any consumer

must travel, falls as new outlets are constructed within a market. Quantity demanded

in the market increases even if the retail price p remains unchanged. The industry

emphasis on market penetration is simply a recognition that r is a function of the

number of outlets in the market, n, as reflected in a recent Forbes Magazine article:

The more stores McDonald's puts in a city, the greater the overall number of trans-actions per capita in that market. Put another way, Greenberg's Law [after JackGreenberg, McDonald's USA chairman] holds that the number of per capita trans-actions varies proportionately with penetration in a market (Samuels, p. 47).

Because the price consumers incur is P = p + cr, reducing r by increasing n is (to the

consumer) tantamount to a reduction in the menu price p. This effect can hardly be

ignored, because r has surely fallen significantly: the number of retail fast food outlets

60 July 2001


Full Price

PO + cr2

PO + crl

Retail Price (Po)

Q2 Q1 Qo Quantity

Figure 1. Quantity consumed as a function of retail priceplus transportation costs

in the United States increased nearly 78% between 1977 and 1992 [U.S. Department of

Commerce (USDC), Census of Retail Trade]. Penetration into previously untapped (e.g.,rural) markets can be viewed as an extreme reduction in travel costs (distance)-even

consumers in the hinterlands could travel to an established market area if absolutely

necessary. The distance consumers are willing to travel is what defines a retail boundary.

Travel distance for the average consumer (r) is not directly measurable, but given r =

f(n), in an empirical analysis a measure can be based on n.2 We use the number of

outlets per square mile (outlet density) and its square, which not only represents the

first two terms in a Taylor series approximation of r, but also allows for diminishing

returns to outlet density. Capturing possible diminishing returns is important given

long-held industry concerns over saturation of markets (e.g., Emerson).

Travel cost per unit of distance (c in the above analysis) corresponds to the opportunity

cost of time in a traditional household production model. In our model, opportunity costs

of time are equally important; it is the value of the time spent traveling that is a primary

determinant of the consumer's willingness to travel a given distance. Of course, there are

also directly observable costs of travel for an individual, such as depreciation of an auto-

mobile, fuel expenses, and repairs, but these are unlikely to play a significant role in the

overall decision to travel to a fast food outlet. The cost of travel, from an opportunity-

cost-of-time perspective, depends on such factors as traffic congestion, road quality,

availability of sidewalks, and other determinants that can affect the speed and ease with

which a given distance can be covered.

2 The problem of determining the distance for a consumer to the nearest outlet is formally equivalent to the problem ofdetermining the price paid for a good by a price searcher observing n prices. This problem was introduced by Stigler in hisseminal work on the economics of information.


,)


The relation between time costs and consumer demand has been studied in othercontexts. DeVany found retail outlet density and outlet size affect demand throughchanges in consumers' expectations of time waiting in line. By increasing firm capacity,or outlet density, demand will increase because of the decreased likelihood the consumerwill experience the cost of waiting in a long queue. Decreasing this uncertainty increasesthe willingness of consumers to travel to a retail outlet, because the expected value oftheir total time costs (travel time plus time spent waiting in line) is reduced.

As demonstrated by Baumol and Ide, the breadth of product offerings by a firm canhave a similar effect on willingness to travel. They note increased variety will, up to apoint, increase the willingness of consumers to travel further distances by increasing theprobability that all their needs are met by a single location. Firms will add products ormenu items until the marginal benefit from increased sales equals the marginal cost ofproviding these additional items. Of course, in a fast food setting, the additional itemsmust not compromise the simplicity of the menu, and the speed and convenience withwhich purchases can be made-so "value meals" and "order-by-number" options remainat the center of even the most extensive fast food menus.

The increase in individual demand owing to lower transport costs (or possibly lowersearch costs) and the consequent "shift" in aggregate demand has been labeled "supplier-induced demand." The premise is that firm location strategy can directly affect consumertransportation costs, and therefore quantity demanded. Supplier-induced demand hasbeen investigated by several researchers, often in the context of healthcare services (e.g.,Newhouse; Evans, Parish, and Sully; Anderson, House, and Ormiston). For a generaltreatment of this topic, interested readers are referred to Benson.

Model Specification

To explore the relationship between outlet penetration and per capita fast food sales, weconstruct a demand model to control for various measures of travel costs, and estimateit using cross-section data from two distinct time periods between which accessibility tofast food has clearly increased. We examine U.S. Department of Commerce/Bureau of theCensus data from 1982 and 1992, the most recent years for which a complete set of datais available from both the Census of Retail Trade (conducted in five-year intervals) andthe decennial Census (years 1980 and 1990). Fortuitously, the 1980s and early 1990s werealso periods of rapid growth for the domestic fast food industry, making these especiallyinteresting periods to examine.

The demand equation was formulated as follows:

(1) Per Capita Fast Food Consumption = f(market characteristics,prices and income, demographics, regional indicators).

Equation (1) was estimated for both 1982 and 1992 using the same cross-section of 85metropolitan areas. Measures of time costs, or availability, are included under the "mar-ket characteristics" variables. Each of the variables are described more fully below.

Data Description

All data are at the Metropolitan Statistical Area (MSA) level, and are readily availablefrom public sources. Most of the variables were compiled from the USA Counties, 1996

62 July 2001


Table 1. Means and Standard Deviations of Variables, 1982 and 1992 (N = 85U.S. Metropolitan Statistical Areas)

1982 1992

Variable Units Mean Std. Dev. Mean Std. Dev.

Dependent Variable:Fast Food Consumption per Capita (sales/price)/population 87.19 25.05 117.20 26.69

Market Characteristic Variables:· Fast Food Outlet Density outlets/100 sq. mi. 12.61 7.50 20.23 11.89· Inexpensive Restaurant Density outlets/100 sq. mi. 7.53 4.58 11.01 6.12· Expensive Restaurant Density outlets/100 sq. mi. 4.22 2.90 6.94 7.17· Food Store Density outlets/100 sq. mi. 18.50 12.22 17.99 12.37· Population Density 100 people/sq. mi. 2.57 1.58 2.85 1.77· Gasoline Consumption per Capita 100 gallons 3.82 0.83 5.60 1.10Price and Income Variables:· Fast Food Price index (price/mean) 78.25 4.16 86.02 3.69· Grocery Price Index index 98.61 4.81 69.76 3.98· Inexpensive Restaurant Price $ 3.41 0.11 3.35 0.26· Expensive Restaurant Price $ 9.15 0.76 8.84 1.32· Income per Capita $000s 11.40 1.63 14.08 1.75Demographic Variables:· Female Labor Force Participation % 51.99 4.69 58.50 4.79

African-American % 10.42 9.94 10.96 10.25· Hispanic % 6.08 13.00 7.38 14.31· Median Age years 28.90 2.02 32.05 2.09· Population Aged 5-9 Years % 7.54 0.87 7.49 0.85· Population Aged 18-21 Years % 6.41 1.82 5.12 1.63· Population Over 65 Years % 9.93 1.95 11.41 2.04· College Graduate % 9.83 2.65 13.05 3.37· Residents per Household number 2.83 0.16 2.69 0.17

Note: 1992 prices and income are adjusted to 1982 levels.

CD-ROM, the Statistical Abstract of the United States, or the Census of Retail Trade (allof which are available through the USDC/Bureau of the Census). All price data are fromthe American Chamber of Commerce Research Association (ACCRA). Any MSA for whichACCRA data were not available in each year was discarded, leaving 85 MSAs for thesample, or 170 observations covering the two years considered. Table 1 reports the 1982and 1992 means and standard deviations for each variable described below.

The dependent variable is a measure of annual per capita fast food consumption(quantity) by MSA in each time period. This measure was constructed by dividing fastfood expenditures [labeled "refreshment places," a subcategory of Standard IndustrialClassification (SIC) 5812 in the Census of Retail Trade] by a price estimate (describedbelow) and population in each MSA.

Market Characteristic Variables

Market characteristic variables are those that control for consumer accessibility tofast food (and its substitutes) across markets. As noted above, we use fast food outletdensity and its square as a proxy for average distance traveled (r in the above analysis).Substitutes for fast food include food prepared at home and meals from table service



restaurants, so we also include outlet density of retail food stores (SIC 54) and tableservice restaurants (subcategory of SIC 5812). Table service restaurants are segmentedinto two price levels: (a) inexpensive restaurants, which are those with an average checksize of under $5 in 1982 and under $7 (nominal) in 1992, and (b) expensive restaurants,which include all those with higher average check sizes. For ease of interpretation, theoutlet density variables are scaled to equal the number of outlets per 100 square miles.

Table service restaurants might compete with fast food on the basis of convenience.Consequently, we expect negative signs for the restaurant density coefficients with theinexpensive segment showing the stronger effect. It is much less likely that food storescompete directly with fast food on the basis of availability, because the cost associatedwith consuming meals at home is largely due to meal preparation, not the purchase ofmeal ingredients. Most consumers necessarily make regular trips to the grocery store toreplenish their home inventory of raw materials used for meal preparation. Althoughease of access to grocery stores is therefore unlikely to affect fast food consumption, westill expect the ceteris paribus effect to be negative.

Moreover, the trend in the grocery industry has been toward larger stores with greaterproduct offerings, thereby increasing the likelihood that all of a consumer's grocery needscan be filled at a single location (Messinger and Chakravarthi). Larger stores explain theslight decrease in the food store density variable between 1982 and 1992 (table 1). Theselarge store formats increasingly offer expanded deli selections and prepared meals fortake-out, which is one way supermarkets have chosen to compete with fast food outlets.That there has been a long-standing concern in the grocery industry over this compe-tition is evident. In 1992, according to Progressive Grocer's 59th Annual Report of theGrocery Industry, 20% of supermarket managers rated the threat from fast food competi-tion as "great," up from 14% in 1982.

Two variables are included as proxies for the consumer cost per unit of travel distance(c in the above analysis): gasoline consumption per capita and population density. Gaso-line consumption is a measure of automobile use. Assuming travel by car is a normalgood, when costs per unit of distance are low, automobile use (and hence gasolineconsumption) will be high. Automobile use is clearly important to the fast food industry,especially given the popularity of drive-thru windows and take-out food. Consequently,greater use of the automobile is expected to positively correlate with per capita consump-tion of fast food. Retail sales of gasoline service stations (from the Census of Retail Trade,SIC 554) divided by the respective average price of regular unleaded fuel in each MSA(reported by ACCRA) and population is our measure of gasoline consumption per capita.

Population density (hundreds of people per square mile) measures the relative easeof mobility within a city: the more densely populated an area, the more burdensome andtime consuming is travel over any given distance, especially by automobile. We hypothe-size a negative correlation between this variable and fast food consumption.

Price and Income Variables3

Retail fast food prices across MSAs are based on ACCRA data. In its Cost of LivingIndex, ACCRA reports city prices for three specific types of fast food meals:

3To facilitate comparison of the parameter estimates across the two years, the 1992 values of all price and income variablesare deflated to 1982 levels (using the All Urban Consumer Price Index).

64 July 2001


* Hamburger Sandwich. Defined as a 1/4-pound patty with cheese, pickle, onion,mustard, and ketchup; McDonald's Quarter Pounder with cheese where available.

* Pizza. Defined as a 12"-13" thin crust regular cheese pizza; Pizza Hut whereavailable.

* Fried Chicken. Defined as a thigh and drumstick-with or without extras, which-ever is lower cost; Kentucky Fried Chicken or Church's where available.

Hamburger, pizza, and chicken outlets represent the three largest sales categories ofthe fast food industry.4 The 1992 Census of Retail Trade reports state-level sales for eachof these types of establishments (along with several other categories), which we used todevelop weights reflecting the percentage of fast food sales held by each type of establish-ment. These weights were used with the ACCRA data in calculating an average fast foodprice for each MSA. Fast food sales divided by this price estimate and population providesthe measure of per capita fast food consumption used as the dependent variable. Thisfast food price (deflated in 1992) also was included as a right-hand-side variable.

The ACCRA Grocery Price Index serves as the price of the food-at-home substitute forfast food. This index is computed quarterly by ACCRA based on a weighted average of27 grocery items believed to represent the typical diet of mid-management executivehouseholds. Data are collected by local Chambers of Commerce, and the cost of livingindices are reported as a service for people who might be relocating to the area.

Relative prices at expensive and inexpensive restaurants were calculated using the"average cost per meal" data in the Census of Retail Trade, Miscellaneous SubjectsSeries. The percentage of restaurants in the appropriate average cost-per-meal pricecategories (defined above in the "market characteristics" section), multiplied by themidpoint of each category and summed, provides a weighted average price for both inex-pensive and expensive restaurants by MSA. ACCRA reports no price information forthese categories. The prices at inexpensive restaurants are positively correlated withthose at fast food establishments, suggesting these may be close substitutes. The extentto which inexpensive restaurants and fast food outlets compete based on price, asopposed to convenience or accessibility, can be inferred from the estimated cross-priceeffects. Annual income in each MSA is measured per capita, using U.S. Department ofCommerce/Bureau of the Census figures.

Demographic Variables

We include the female labor force participation rate of each MSA as a measure of theopportunity cost of time. Previous studies using panel data have generally reportedconsumption of food away from home is greater in households where the female familymember works outside the home.

Four variables are used to control for the age distribution of each MSA: median ageof the population, and the percentage of the population falling within each of threeseparate age groups (5-9, 18-21, and over 65 years). The 5-9 age group was includedbecause of the substantial marketing efforts aimed toward children, especially by the

4 In 1992, national sales from fast food hamburger, pizza, and chicken outlets were roughly 43.6%, 15%, and 8.8%, respec-tively, of U.S. total fast food sales (USDC, Census of Retail Trade).



largest participants in the industry. The 18-21 age group is widely regarded to representthe "heavy users" of fast food, while the 65 and older age group is distinguished asconsuming the fewest fast food meals (Emerson).

The MSA population divided by the respective number of households captures theeffect of household size, which previous studies (Redman; McCracken and Brandt;Manchester) have found to have a negative effect on household (and hence per capita)consumption of FAFH. The percentage of the population with at least a four-year collegedegree is included to explore whether education affects preferences for fast food. Theexpected effects of higher education on fast food consumption are not clear. Two racialcomposition variables are included to control for possible preference variations based onculture: the percentage of the population classified as African-American, and the percent-age classified as Hispanic-the two largest minority segments of the population.

Finally, dummy variables are included corresponding to eight geographic regions ofthe United States.5 These variables are included simply to control for the possible exist-ence of distinct food consumption patterns across regions. Such patterns could be theresult of differences in fast food preferences, pricing patterns, promotion strategies, orother factors that might exhibit regional variation. The metropolitan areas included ineach region are presented in appendix table Al.

Fast Food Demand Estimates, 1982 and 1992

The two cross-section demand equations (one for 1982 and the other for 1992) were esti-mated together as a set of seemingly unrelated regression (SUR) equations.6 Controllingfor cross-equation correlation in this manner not only improves efficiency, but alsoensures valid tests of parameter estimates across equations.7 The estimated correlationbetween the two equations, i.e., the contemporaneous correlation between the error termsacross equations, was 0.454.8 The parameter estimates for each year are presented intable 2, and F-tests of the joint significance of selected groups of variables are presentedin table 3.

Market Characteristic Variables

Given our focus on accessibility, the most important results are the parameter estimatesfor the fast food outlet density variable (table 2). In both years, fast food outlet densityis positive and significant, strong evidence of a direct link between the number of outlets

5 Because New England has only a single observation in this data set (Hartford, Connecticut), it was grouped with theMiddle Atlantic region, leaving eight separate regions.

6 A Chow test of the null hypothesis of whether the two years can be pooled was calculated to test for a structural changein demand over time. With an intercept shifter in the pooled model distinguishing between years, the Chow test result wasF28,114 = 1.89, which rejects the null at a = 0.05. Removing the intercept shifter only strengthens this result.

7 Note that this procedure is different from the usual application of SUR in demand analysis, where it is typically used tocapture correlation across cross-sectional units, such as commodity demands, markets, or industries. This application couldbe viewed as a time-series, autocorrelation correction across the two periods, where the correlation is allowed to vary over time,but not across cross-sectional observations.

8 If the correlation between the two equations is zero, i.e., the variance-covariance matrix of the system of equations isdiagonal, SUR would provide no efficiency gains and estimation would simply revert to OLS. A Lagrange multiplier test ofwhether the variance-covariance matrix is diagonal produces X1 = 17.5, allowing us to reject the null hypothesis of zero corre-lation between the two equations.

66 July 2001


Table 2. SUR Parameter Estimates of Fast Food Consumption, 1982 and 1992

1982 1992

Parameter ParameterVariable Estimate t-Value Estimate t-Value

Intercept 25.214 0.14 155.261 0.63Market Characteristic Variables:· Fast Food Outlet Density 3.201*** 2.91 3.158*** 3.62· (Fast Food Outlet Density) 2 -0.031 -1.58 0.004 0.41· Inexpensive Restaurant Density 0.827 1.17 -0.248 -0.30· Expensive Restaurant Density - 1.903* -1.87 -0.352 -0.51· Food Store Density -0.385 -0.72 -0.366 -0.63· Population Density -8.507 -1.65 -20.981*** -2.82* Gasoline Consumption per Capita 2.559 1.09 - 1.145 -0.72Price and Income Variables:· Fast Food Price -1.138*** -2.85 -2.539*** -4.23· Grocery Price Index 0.307 0.91 0.573 1.00· Inexpensive Restaurant Price 5.691 0.46 -5.429 -0.86* Expensive Restaurant Price 2.955 1.29 -0.784 -0.65* Income per Capita 5.756** 2.45 3.214 1.58Demographic Variables:· Female Labor Force Participation -0.613 -1.21 0.375 0.54· African-American 0.211 0.91 0.580** 2.55· Hispanic 0.262 1.27 0.702*** 2.63· Median Age 1.130 0.37 3.113 0.71

Population Aged 5-9 Years 7.068 1.19 - 1.605 -0.26Population Aged 18-21 Years 7.545*** 2.68 3.578 0.90

· Population Over 65 Years -0.351 -0.55 -0.769 -0.85· College Graduate 0.425 0.35 -0.773 -0.76* Residents per Household -43.601* -1.76 3.241 0.11Regional Indicators:· New England/Middle Atlantic 4.200 0.41 -2.404 -0.23· South Atlantic 40.235*** 3.70 37.984*** 3.55* East North Central 16.358* 1.89 13.516 1.41· East South Central 33.819*** 3.29 41.853*** 3.84· West North Central 8.116 0.97 19.180* 1.85* West South Central 26.216*** 3.32 12.361 1.30

Mountain 3.191 0.43 6.749 0.75

System Weighted R2 = 0.821, n = (85 x 2) = 170

Note: Single, double, and triple asterisks (*) denote significance at the 10%, 5%, and 1% levels, respectively (based on a two-tailed test).

Table 3. F-Tests of the Joint Significance of Variable Groups, 1982 and 1992Demand Estimates

EqualVariable Group 1982 1992 Across Periods

Market Characteristics F7,112 = 5.04*** F7112 = 3.87*** F7,112 = 2.38**Prices and Income F 1 12 = 3.56*** F5, 112 = 4.49*** F5, 112 = 2.81**Demographics F, 1 12 = 5.70*** F, 112 = 2.38** F 1 12 = 2.37**Regional Indicators F7112 = 3.95*** F7, 1 12 = 4.29*** F7, 112 = 1.54

Note: Double and triple asterisks (*) denote significance at the 5% and 1% levels, respectively (based on a two-tailed test).



serving a market and per capita fast food consumption. This result supports the hypoth-esis that consumer accessibility to fast food contributes directly to its convenience.Density-squared is not significant at standard levels in either equation. A test of whetherthe fast food density and density-squared estimates are equal across periods is rejectedat the 5% level (F2, 1 12 = 3.26), although a test that the density variable alone is equalacross periods is not rejected. This finding reveals, despite a 60% average increase inoutlet density in this sample between 1982 and 1992, the marginal effect of additionaloutlets on consumption remained strongly positive. Through 1992, market saturationwas not a serious threat; firms continued to find locations within these same marketswhich improved consumer accessibility and directly contributed to greater sales.

Neither inexpensive restaurant density nor food store density is significant in eitherequation. Surprisingly, however, the density of expensive restaurants in the MSA isnegative and significant at the 10% level in 1982. We know these goods are poor substi-tutes; the relationship may reflect differences in tastes between areas where expensiverestaurants are prevalent and areas where they are not (perhaps even revealing a mildform of snobbery). The absence of an expensive-restaurant effect in 1992 likely reflectsa greater acceptance of fast food among all consumers. The lack of significance of theother density variables indicates neither restaurants nor food stores compete directlywith fast food on the basis of travel costs. For food stores this is not surprising, given thediscussion in the previous section regarding the grocery industry and the relativeimportance of meal preparation time as opposed to time spent purchasing the inputs forconsumption of food at home. The absence of a significant parameter estimate for theinexpensive restaurant density variable suggests that based on convenience, inexpensiverestaurants are a poor substitute for fast food. This issue is readdressed in the discussionof the price parameter estimates below.

The proxy variables for travel cost (c) give mixed results. Gasoline consumption is notsignificant at standard levels in either equation, but population density has the expectednegative effect and is significant in the 1992 equation. Population density in some waysis a superior measure of travel costs, because it does not focus on a single mode oftransportation. In highly populated areas, such as inner cities, gasoline consumption percapita will naturally be relatively low, because public transportation is more commonand walking may be a viable alternative. As fast food firms expand into inner city localesfrom the suburbs, sales at these outlets are not as directly tied to the suburban roadsystem; hence gasoline consumption would have a negligible effect.

The much stronger negative effect of population density in 1992 suggests a structuralchange in the demand equation. To the extent this variable measures the cost of travel,this finding indicates fast food consumption is adversely affected by the inconvenienceassociated with mobility in densely populated areas. This result implies an increase overtime in the value consumers place on convenience. While it is difficult to develop a trulyaccurate measure of consumer travel costs, these results nevertheless suggest that easeof transportation plays an important role in the demand for fast food. However, traveldistance as proxied by outlet density has a stronger, more obvious and direct effect.

Ajoint F-test reveals the market characteristic variables as a group contribute signifi-cant explanatory power in each period. This simply provides credence to the argumentthat consumption decisions are driven by more than just retail price. Especially forproducts capitalizing on the demand for convenience, retail outlet density appears to bea useful measure for capturing the full price to the consumer. A test of the null hypothesis

68 July 2001


that this group of variables has the same effect in each time period is rejected (F7 112 =

2.38,p-value = 0.03), suggesting some overall change in the marginal effect of outlet den-sity over time. Given the dynamics of the restaurant and fast food industries, especiallyin regard to outlet proliferation strategies, marketing-induced changes in preferences,and ever-increasing opportunity costs of time for consumers, it is not surprising that themarginal effect of outlet density was not constant over this 10-year period.

Price and Income Variables

With the expected signs, fast food price is significant in both periods, but per capita in-come is statistically significant only in 1982. None of the substitute prices are significant.A test of whether the fast food price parameter estimate is equal across periods suggestssensitivity to price has increased over time (F1, 112 = 4.05, p-value = 0.05). We wouldexpect sensitivity to retail price to increase as availability increases because travel costshave been reduced, making the retail price a larger proportion of the total price. (Thiscontention also is supported in the price elasticity estimates in table 4.) The increase inelasticity in 1992 (over 1982) could reflect the growth over time in the number of con-venient meal choices, especially microwavable foods and prepared meals (home mealreplacements) offered by grocery stores.

An increase in the number of substitute products in a market is typically associatedwith an increase in own-price elasticity. According to the Statistical Abstract of theUnited States (USDC), U.S. households owning a microwave oven increased from20.7% in 1982, to 81% in 1993. Clearly, household adoption of this technology has ledto a decline over time in the burden of preparing meals at home, to the likely dismayof the fast food industry. The ACCRA Grocery Price Index would not capture thiseffect as it does not measure the convenience of grocery items, which is the basis onwhich home meal replacements and microwavable foods most likely compete withfast food.

The fact that neither price nor availability of inexpensive restaurants is significant ineither equation is not necessarily surprising: it corroborates the hypothesis that conven-ience drives the demand for fast food. Few other FAFH options are able to compete inthis arena. Inexpensive restaurants can offer meals of equal or greater quality to fastfood for about the same retail price, but if a high value is placed on time, fast food isvastly cheaper. At current levels, neither differences in price between table servicerestaurants and fast food, nor differences in accessibility will have a noticeable effect onfast food sales because the time cost of acquiring a meal is always greater at a tableservice restaurant. Even when the retail price is similar to fast food, the products, whenviewed from a convenience perspective, are not. The apparent lack of competitionbetween inexpensive restaurants and fast food outlets suggests the expansion of the fastfood industry is not necessarily a threat to the inexpensive restaurant segment, sincedifferent factors drive the demand for these products. Accordingly, small, family-owned,inexpensive restaurants continue to be viable business opportunities in many markets(National Restaurant Association 1995).

Regarding income, a test of whether the income parameter estimate is equal acrossperiods is not rejected (F1 ,122 = 0.83), implying the income response has not changedgreatly over time, despite a noticeable difference in the parameter estimate. The smallerbut insignificant parameter estimate, along with a rise in average income, resulted



Table 4. Fast Food Price and Income per Capita Elasticity Estimates, 1982and 1992 (at data means)

EqualDescription 1982 1992 Across Periods

Fast Food Price -1.022*** -1.884*** F, 112 = 4.05**

Income per Capita 0.753** 0.386 F1,1 2 = 0.83

Note: Double and triple asterisks (*) denote significance at the 5% and 1% levels, respectively (based on a two-tailed test).

in a decrease in the income elasticity estimate from 0.753 in 1982, to 0.386 in 1992

(table 4). Most other studies have found FAFH consumption has an inelastic income

response.

Demographic Variables

As a group, the demographic variables contribute significant explanatory power eachyear (table 3), but individually most are not significant or have estimated effects thatwere not anticipated. The only statistically significant age variable is the percentageof the population in the 18-21 age group in 1982, which displays the expectedrelationship (table 2). In 1982, there also was a tendency for large households to

consume fewer fast food meals. This result was expected, because meal preparation

at home becomes more economical as family size increases. By 1992, neither of these

relationships remained significant. A test of the null hypothesis that the group of

demographic variable coefficients are equal across time is rejected at the 5% level

(F9,1 12 = 2.37, p-value = 0.02).Both of the racial composition variables are significant and positive in 1992. Many

studies find significant differences in eating patterns based on race, including studiesfocusing on FAFH (e.g., Prochaska and Schrimper; Redman). Typically, these findingsare attributed to differences in tastes and preferences across racial classes. We find astrengthening effect of race on fast food consumption over time, which could be indicativeof a growing trend for fast food firms to expand into areas with high African-Americanand Hispanic populations, thereby increasing the availability to these consumers andallowing any differences in preferences to become more noticeable.

The percentage of female labor force participation is not significant in either of theperiods. This reveals a possible weakness of market-level data. In panel data studies, theemployment status of females is typically captured with a dummy variable, explicitlycomparing the consumption behavior of working and nonworking females. Our datareport the percentage of the total MSA female population in the workforce, making noexplicit comparison between those already in the workforce and those who are not. It isa truism that working females have a higher opportunity cost of time, and therefore theyshould value the convenience of fast food. This value is not detectable in our market-level

data, perhaps due to an aggregation effect. Confounding factors with female workforceparticipation also may exist. For example, in some households females earn a secondincome, whereas in other households single mothers must support children by workinglow-paying jobs.

70 July 2001


Regional Indicators

Finally, as a group, the regional indicators are significant in each time period (table 3),

providing evidence of differing consumption patterns across regions. The omitted region

is the Pacific. In both years, the southern regions (South Atlantic, East South Central,

and West South Central) tend to consume the largest quantities of fast food, while con-

sumers in the New England/Middle Atlantic and Pacific regions consume the least. A test

that the regional effects are the same in both periods is not rejected at standard levels

(F7, 112 = 1.54, p-value = 0.16).

Are Fast Food Price and OutletDensity Exogenous?

Using arguments founded in household production theory, we have haillustrated analytically

why we would expect the number of fast food outlets in a market to positively influence

sales. Our empirical analysis appears to confirm this hypothesis, implying that much of

the observed growth in consumption is actually the result of increased convenience and

availability of fast food rather than a change in the nature of demand. The final issue we

address is whether outlet density and fast food price are exogenous in our model.

A fundamental question is whether firms increase their presence in markets where

consumption is already high or adjust prices accordingly, rendering these variables

endogenous. Alternatively, outlet penetration may itself be used as part of a sales growth

strategy. Note this does not rule out the possibility that location decisions or prices could

be influenced by other explanatory variables in the model, such as demographics. Exog-

eneity cannot be identified per se, but we can determine if the variables in question are

uncorrelated with the disturbance. This would provide evidence in favor of exogeneity,

and assurance our estimates are consistent.A Durbin-Wu-Hausman procedure (Davidson and MacKinnon) was used to test for

endogeneity of fast food price and fast food density. The instruments used in the test

procedure included all of the predetermined variables from the original model, plus

measures of wage rates, rent, and building maintenance costs, the most important cost

factors for fast food.9 Table 5 presents the t- and F-statistics from the predicted reduced-form coefficients. The first two rows report results where price and density were tested

independently, and the third and fourth rows show results of a joint test.

Consistent with the industry view on outlet proliferation as a marketing strategy, test

statistic values do not indicate the presence of endogeneity. Furthermore, if market-level

sales were a reasonable indicator of consumer demand for additional services and new

capacity, firms might be expected to increase the size of existing units, possibly capturing

economies of scale, and avoiding the large fixed costs of constructing new outlets. This

is clearly not the case, as outlets have on average become smaller over time: the number

of seats per U.S. fast food outlet has decreased from an average of 51.2 in 1982, to 48.1

9 Average wage rates were calculated as the total retail payroll divided by the number of retail employees, rent expense wasapproximated with the average residential apartment rent (under the assumption that residential and commercial rent costsare highly correlated), and maintenance costs were approximated with the average annual cost of upkeep, including mortgagepayments, for residential properties (also under the assumption that residential maintenance costs and commercial mainten-ance costs are highly correlated). All of these variables are readily available from the U.S. Department of Commerce/Bureauof the Census.


Journal of Agricultural and Resource Economics

Table 5. Exogeneity Tests of Fast Food Price and Fast Food Outlet Density,1982 and 1992 t- and F-Values

Description 1982 1992

Independent t-Tests: Fast Food Price 1.492 -0.984Fast Food Outlet Density -1.140 1.218

Joint Tests: Fast Food Price 0.659 - 0.952

Fast Food Outlet Density -0.155 1.358

F-Test F2,54 = 1.18 F2, 54 = 1.12

Notes: The null hypothesis is that there is no correlation between the variables in question and the error. None of the teststatistics are significant at standard levels. The t-statistics are from independent tests; the F-statistics test the jointsignificance of fast food price and fast food outlet density within the same equation.

in 1992 (USDC, Census of Retail Trade). The endogeneity test provides evidence in sup-port of our original assertion that growth in the number of outlets is more a cause of,rather than a response to, consumer demand for fast food.

Conclusions

The demand for fast food depends heavily on the demand for convenience, and an impor-tant component of convenience is the ease of consumer access to the product. An increasein the number of fast food outlets in a market directly increases quantity consumed bydecreasing the costs of obtaining a fast food meal. Proliferation of fast food outletsexplains much of the historic long-term growth of this industry. The fast food industryhas continually found ways to make its products more accessible, and this effortcontinues today, with retail outlets appearing in such varied locations as office buildings,department stores, and airports. Failing to account for this changing structure of themarketplace could lead to the inference that over time there have been dramatic changesin tastes, or that the continued growth in fast food has resulted solely from a heightenedconsumer demand. Our results strongly suggest much of this growth in consumption isattributable to an increasing supply of convenience.

We did not find any empirical evidence to indicate fast food firms are in direct compe-tition with table service restaurants in an equivalent price range. This finding supportsthe common observation that few table service restaurants offer the same level of time-saving convenience as fast food, and it is the convenience-supplying attributes that givefast food its primary appeal. We did find evidence of an increase in the own-priceelasticity of demand for fast food over time. Increased sensitivity to menu prices occursbecause, as distance traveled to fast food outlets declines, menu prices account for alarger portion of the "full" price of fast food.

Except for a growing tendency for Black and Hispanic consumers to have higher thanaverage fast food consumption, we found little evidence to suggest consumption of fastfood varies across demographic groups or income levels. However, significant differencesin regional consumption of fast food have persisted over time. With the broad appeal ofthis product extending across most segments of the population, if firms maintain theirefforts to increase the availability of fast food, industry growth is expected to continue.

[Received May 2000; final revision received March 2001.]

72 July 2001


References

American Chamber of Commerce Research Association. Cost of Living Index and Grocery Price Index.

ACCRA, Alexandria VA. Various years, 1982-92.Anderson, R. K., D. House, and M. B. Ormiston. "A Theory of PhysicianBehavior with Supplier Induced

Demand." S. Econ. J. 48,1(1981):124-33.Baumol, W. J., and E. A. Ide. "Variety in Retailing." Manage. Sci. 3,1(1956):93-101.

Becker, G. S. "A Theory of the Allocation of Time." Econ. J. 75(1965):493-517.

Benson, B. "The Impact of Entry at a Distance on Market Demand." Rev. Regional Stud. 10,2(1980):

62-68.Brown, D. "The Restaurant and Fast Food Race: Who's Winning?" S. Econ. J. 56,4(1990):984-95.

Byrne, P. J., O. Capps, Jr., and A. Saha. "Analysis of Food Away from Home Expenditure Patterns for

U.S. Households." Amer. J. Agr. Econ. 78,3(1996):614-27.."Analysis of Quick Serve, Mid-Scale, and Up-Scale Food Away from Home Expenditures."

Internat. Food and Agribus. Manage. Rev. 1(1998):51-72.Crafton, S. M. "Convenience Store Pricing and the Value of Time: A Note on the Becker-Devany Full

Price Model." S. Econ. J. 45,4(1979):1254-60.Davidson, R., and J. G. MacKinnon. Estimation and Inference in Econometrics. New York: Oxford Uni-

versity Press, 1993.DeVany, A. S. "Uncertainty, Waiting Time, and Capacity Utilization: A Stochastic Theory of Product

Quality." J. Polit. Econ. 84,3(1983):523-41.Ekelund, R. B., and J. K. Watson. "Restaurant Cuisine, Fast Food, and Ethnic Edibles: An Empirical

Note on Household Meal Production." Kyklos 44,4(1991):613-27.Emerson, R. L. The New Economics of Fast Food. New York: Van Nostrand-Reinhold Publishing, 1990.

Evans, R. G., M. A. Parish, and F. Sully. "Medical Productivity, Scale Effects, and Demand Generation."

Can. J. Econ. 6,3(1973):376-93.Hausman, J. A. "Specification Tests in Econometrics." Econometrica 46,5(1978):1251-72.

Hiemstra, S. J., and W. G. Kim. "Factors Affecting Expenditures for Food Away from Home in Commer-

cial Establishments by Type of Eating Place and Meal Occasion." Hospitality Res. J. 19,3(1995):15-31.

Kinsey, J. "Working Wives and the Marginal Propensity to Consume Food Away from Home." Amer. J.

Agr. Econ. 65,1(1983):10-19.Lee, J., and M. Brown. "Food Expenditures at Home and Away from Home in the United States: A

Switching Regression Analysis." Rev. Econ. and Statis. 68(1986):142-47.Manchester, A. "U.S. Food Spending and Income, Changes Through the Years." Agr. Info. Bull. No. 618,

USDA/Economic Research Service, Washington DC, January 1991.

Manchester, A., and A. Clauson. "1994 Spending for Food Away from Home Outpaces Food at Home."

In Food Review, pp. 12-15. USDA/Economic Research Service, Washington DC, July 1995 (and subse-

quent updates).McCracken, V. A., and J. A. Brandt. "Household Consumption of Food-Away-from-Home: Total Expendi-

ture and by Type of Food Facility." Amer. J. Agr. Econ. 69,2(1987):274-84.McDonald's Corporation USA. The McDonald's Corporation Annual Report. Annual report to share-

holders. McDonald's USA, Oak Brook IL, 1994.Messinger, P. R., and N. Chakravarthi. "A Model of Retail Formats Based on Consumers' Economizing

on Shopping Time." Marketing Sci. 16(1997):1-23.National Restaurant Association. Restaurants USA. NRA monthly publication, Washington DC. Various

issues, 1995-96.Newhouse, J. P. "A Model of Physician Pricing." S. Econ. J. 37,1(1970):174-83.

Prochaska, F. J., and R. A. Schrimper. "Opportunity Cost and Other Socioeconomic Effects on Away-

from-Home Food Consumption." Amer. J. Agr. Econ. 55,3(1973):595-603.Progressive Grocer [staffl. Progressive Grocer's 49th Annual Report of the Grocery Industry. Manhattan/

New York, April 1982.Progressive Grocer's 59th Annual Report of the Grocery Industry. Manhattan/New York, April

1992.Redman, B. J. "The Impact of Women's Time Allocation on Expenditure for Meals Away from Home and

Prepared Foods." Amer. J. Agr. Econ. 62,2(1980):234-37.

Jekanowski, Binkley, antd Eales


Samuels, G. "Golden Arches Galore." Forbes Magazine (4 November 1996):46-48.Sexauer, B. "The Effects of Demographic Shifts and Changes in the Income Distribution of Food-Away-

from-Home Expenditure." Amer. J. Agr. Econ. 61,3(1979):1046-57.Stigler, G. J. "The Economics of Information." J. Polit. Econ. 69(1961):213-25.U.S. Department of Agriculture, Economic Research Service, Food Expenditures Briefing Room. Food

Expenditure Tables, 1967-1999. USDA/ERS, Washington DC. Online. Available at http://www.ers.usda.gov/briefing/CPIFoodAndExpenditures/Data/foodexpendituretables.htm.

U.S. Department of Commerce, Bureau of the Census. Census of Retail Trade, Miscellaneous SubjectsSeries. USDC/Bureau of the Census, Washington DC, 1982, 1992.

. Statistical Abstract of the United States. USDC/Bureau of the Census, Washington DC. Variousyears, 1982-92.

. USA Counties, 1996. CD-ROM, USDC/Bureau of the Census, Washington DC, 1996.Yen, S. T. "Working Wives and Food Away from Home: The Box-Cox Double-Hurdle Model." Amer. J.

Agr. Econ. 75,4(1993):884-95.

Appendix

Table Al. 85 Metropolitan Areas Examined, Distributed into Eight Census Regions

New England/Middle Atlantic: a

Hartford, CTBinghamton, NYHarrisburg, PALancaster, PASyracuse, NYYork, PA

South Atlantic:

Albany, GAAthens, GAAtlanta, GAAugusta, GAColumbus, GACharlotte, NCGreensboro, NCHickory, NCRaleigh, NCColumbia, SCGainesville, FLMiami, FLRoanoke, VAHuntington, WV

Pacific:

Bakersfield, CALos Angeles, CASan Diego, CAPortland, ORRichland, WASeattle, WA

East North Central:

Champaign, ILDecatur, ILRockford, ILSpringfield, ILFort Wayne, INIndianapolis, INSouth Bend, INLansing, MICincinnati, OHCleveland, OHColumbus, OHYoungstown, OHDayton, OHAppleton, WIGreen Bay, WIJanesville, WILa Crosse, WIWausau, WI

West North Central:

Cedar Rapids, IAIowa City, IAWichita, KSMinneapolis, MNRochester, MNColumbia, MOKansas City, MOSt. Joseph, MOSt. Louis, MOSpringfield, MOLincoln, NEOmaha, NE

East South Central:

Birmingham, ALHuntsville, ALMontgomery, ALLexington, KYLouisville, KYChattanooga, TNKnoxville, TNMemphis, TNNashville, TN

West South Central:

New Orleans, LAOklahoma City, OKAmarillo, TXBeaumont, TXDallas, TXEl Paso, TXHouston, TXLubbock, TXMcAllen, TXSan Antonio, TXWaco, TX

Mountain:

Phoenix, AZColorado Springs, CODenver, COBoise City, IDAlbuquerque, NMLas Vegas, NVReno, NVBillings, MTSalt Lake City, UT

a Due to data limitations, New England and Middle Atlantic are combined into a single region.

74 July 2001

convenience, accessibility, and the demand for fast food · convenience, accessibility, and the...

Documents